FP16Benchmark: Allow fp32 comparison using cblas (#56)

Summary:
FP16Benchmark: Allow comparison against fp32
using any local cblas library if MKL not found.
Pull Request resolved: https://github.com/pytorch/FBGEMM/pull/56

Reviewed By: jianyuh

Differential Revision: D13645545

Pulled By: dskhudia

fbshipit-source-id: ca98e84bfb85eb3b0edebad664d211c3af8db309

3 files changed, 33 insertions, 7 deletions

diff --git a/.gitignore b/.gitignore
new file mode 100644
index 0000000..567609b
--- /dev/null
+++ b/.gitignore
@@ -0,0 +1 @@
+build/
diff --git a/bench/CMakeLists.txt b/bench/CMakeLists.txt
index cc07014..a58eed7 100644
--- a/bench/CMakeLists.txt
+++ b/bench/CMakeLists.txt
@@ -1,6 +1,10 @@
 cmake_minimum_required(VERSION 3.5 FATAL_ERROR)
 
 find_package(MKL)
+if (NOT ${MKL_FOUND})
+  find_package(BLAS)
+endif()
+
 #benchmarks
 macro(add_benchmark BENCHNAME)
   add_executable(${BENCHNAME} ${ARGN}
@@ -13,11 +17,18 @@ macro(add_benchmark BENCHNAME)
   target_link_libraries(${BENCHNAME} fbgemm)
   add_dependencies(${BENCHNAME} fbgemm)
   if(${MKL_FOUND})
+    message(STATUS "MKL_LIBRARIES= ${MKL_LIBRARIES}")
     target_include_directories(${BENCHNAME} PRIVATE "${MKL_INCLUDE_DIR}")
     target_link_libraries(${BENCHNAME} "${MKL_LIBRARIES}")
     target_compile_options(${BENCHNAME} PRIVATE
       "-DUSE_MKL")
   endif()
+  if (${BLAS_FOUND})
+    message(STATUS "BLAS_LIBRARIES= ${BLAS_LIBRARIES}")
+    target_compile_options(${BENCHNAME} PRIVATE "-DUSE_BLAS")
+    target_link_libraries(${BENCHNAME} "${BLAS_LIBRARIES}")
+  endif()
+
   set_target_properties(${BENCHNAME} PROPERTIES FOLDER test)
 endmacro()
 
diff --git a/bench/FP16Benchmark.cc b/bench/FP16Benchmark.cc
index c42ee29..efb043d 100644
--- a/bench/FP16Benchmark.cc
+++ b/bench/FP16Benchmark.cc
@@ -12,6 +12,10 @@
 #include <mkl.h>
 #endif
 
+#ifdef USE_BLAS
+#include <cblas.h>
+#endif
+
 #ifdef _OPENMP
 #include <omp.h>
 #endif
@@ -101,7 +105,7 @@ void performance_test() {
     // warm up MKL and fbgemm
     // check correctness at the same time
     for (auto w = 0; w < 3; w++) {
-#ifdef USE_MKL
+#if defined(USE_MKL) || defined(USE_BLAS)
       cblas_sgemm(
           CblasRowMajor,
           CblasNoTrans,
@@ -121,19 +125,29 @@ void performance_test() {
       cblas_gemm_compute(
           matrix_op_t::NoTranspose, m, A.data(), Bp, beta, C_fb.data());
 
-#ifdef USE_MKL
+#if defined(USE_MKL) || defined(USE_BLAS)
       // Compare results
       for (auto i = 0; i < C_ref.size(); i++) {
-        // printf("%f %f\n", C_ref[i], C_fb[i]);
-        assert(std::abs(C_ref[i] - C_fb[i]) < 1e-3);
+        if (std::abs(C_ref[i] - C_fb[i]) > 1e-3) {
+          fprintf(
+              stderr,
+              "Error: too high diff between fp32 ref %f and fp16 %f\n",
+              C_ref[i],
+              C_fb[i]);
+          return;
+        }
       }
 #endif
     }
 
     chrono::time_point<chrono::system_clock> t_begin, t_end;
-#ifdef USE_MKL
+#if defined(USE_MKL) || defined(USE_BLAS)
     // Gold via MKL sgemm
+#if defined(USE_MKL)
     type = "MKL_FP32";
+#else
+    type = "BLAS_FP32";
+#endif
     ttot = 0;
     for (auto it = -3; it < NITER; it++) {
       if (flush) {
@@ -166,7 +180,7 @@ void performance_test() {
     gflops = nflops / ttot / 1e9;
     gbs = nbytes / ttot / 1e9;
     printf(
-        "\n%15s m = %5d n = %5d k = %5d Gflops = %8.4lf GBytes = %8.4lf\n",
+        "\n%30s m = %5d n = %5d k = %5d Gflops = %8.4lf GBytes = %8.4lf\n",
         type.c_str(),
         m,
         n,
@@ -199,7 +213,7 @@ void performance_test() {
     gflops = nflops / ttot / 1e9;
     gbs = nbytes / ttot / 1e9;
     printf(
-        "%15s m = %5d n = %5d k = %5d Gflops = %8.4lf GBytes = %8.4lf\n",
+        "%30s m = %5d n = %5d k = %5d Gflops = %8.4lf GBytes = %8.4lf\n",
         type.c_str(),
         m,
         n,